Milton and Blaber: Sexual maturity and spawning of tuna baitfish in the Solomon Islands 



223 



prey (from those found by Milton et al. 1990) were 

 counted. Spratelloides eggs are demersal (Leis and 

 Trnski 1989) so were not sampled by this method. 



Data analyses 



The proportion of each 1 mm length-class that were sex- 

 ually mature was compared between sites. The normal 

 approximation to the binomial distribution was used to 

 estimate the 95% confidence limits of the proportion 

 mature in any length-class. Confidence limits of the 

 estimated proportion of the mature population spawn- 

 ing each month were calculated in a similar way. 



To assess whether the timing of spawning was ran- 

 dom, we calculated the proportion of the population 

 spawning from the fraction of the entire sample of each 

 species at each site during the study. This proportion 

 was then compared with the proportion spawning each 

 month. Monthly proportions greater than the 95% con- 

 fidence limits to the normal approximation to a bi- 

 nomial distribution were scored as a plus sign. A non- 

 parametric runs test (Sokal and Rohlf 1981) was used 

 to test whether the distribution of plus signs was 

 random. 



The relationship between the proportion of fish in 

 each monthly sample (of each species at each site) that 

 were longer than the mean adult length at that site, 

 and the proportion in that sample that were spawning, 

 was examined using linear regression. A significant 

 positive relationship between these proportions was 

 used to assess whether fish spawned as soon as they 

 were physiologically capable of doing so. 



The relationships between possible proximate stimuli 

 and the proportion spawning were compared by step- 

 wise regression analysis. Variables that improved the 

 fit were added to the model until the best three- variable 

 model was obtained or the most significant fit was 

 found. The proportion of each sample of each species 

 in spawning condition was transformed by its arcsine 

 square-root to reduce possible bias due to an excess of 

 low values (Sokal and Rohlf 1981). The following 11 

 variables were compared for each species at each site: 

 (1) sea-surface temperature, (2,3) prey biomass and 

 density, (4) moon phase, (5) tide range, (6) cloud cover, 

 (7) wind speed, and (8-11) rainfall between samples. 

 Previous studies of spawning by these species (Dalzell 

 1985, 1987b) and other tropical inshore fishes (e.g., 

 Johannes 1978, Walsh 1987) suggested that these vari- 

 ables may be important cues for these species. Initial- 

 ly, salinity and current speed were also measured, but 

 as they varied little, they were not included in the 

 analysis. 



The variable moon phase was calculated by fitting 

 a curve of the form y = sin(2nx) + cos(2nx) where x = 

 number of days since the last full moon prior to the 



start of sampling divided by 29.5 (days in a lunar 

 month). This variable has higher values about the new 

 and full moon. Rainfall data were regressed in several 

 ways to assess the influence they may have on baitfish 

 spawning: total rainfall since previous sample (usually 

 1 month)(Total); number of days since rain (Days); 

 number of days since rain >25mm (Days 25 mm); and 

 number of days of rain since previous sample (Rain). 

 Rainfall, cloud cover, and wind variables were trans- 

 formed by their square root to normalize skewed 

 values. Except for moon phase, positive relationships 

 between the proportion of each species spawning and 

 proximate variables are indicative of greater spawn- 

 ing activity at higher values. 



To assess whether all variables were independent, 

 proximate variables were correlated with one another. 

 Principal component analysis (Sokal and Rohlf 1981) 

 was also used to help identify variables that covaried 

 within and between sites. A subset of proximate vari- 

 ables that behaved independently was identified and 

 analysed separately. Where a potential stimulus had 

 been measured in several ways (e.g., rainfall), the most 

 biologically appropriate was chosen. 



Results 



Physical environment 



The mean sea-surface temperature and monthly rain- 

 fall at each site varied seasonally during the sampling 

 period (Figs. 1-3). The temperature ranged from 27.5° 

 to 32.5°C at all sites, with lower temperatures during 

 the dry season (May-September). Temperatures were 

 lower in 1987 than 1988 at all sites. Rainfall occurred 

 in all months at all sites. The amount of rainfall and 

 its monthly distribution pattern were similar at the two 

 closest sites, Munda and Vona Vona (r 2 0.48, P< 

 0.05), but Tulagi had a higher rainfall and different pat- 

 tern of distribution. 



Maturation 



A total of 1 159 fish of six baitfish species from the three 

 sites were examined histologically, including over 200 

 of each of the four most abundant species: Encrasi- 

 cholina devisi, E. heterolobus, Spratelloides delicatulus, 

 and S. lewisi. 



Encrasichollna Histological examination of ripe ova- 

 ries of E. devisi and E. heterolobus showed that almost 

 all oocytes were in the most advanced stage of develop- 

 ment, with a few at the yolk precursor stage (Table 1) 

 or less developed. Although many females (~15%) of 

 those examined had hydrated eggs, no post-vitellogenic 

 follicles or spent fish (Stage 6) were observed. 



